A device disclosed in patent literature 1 by the applicant of this application has a procedure of obtaining a time-series waveform of a frequency from a time-series waveform of a biological signal mainly indicating the pulsation of a cardiovascular system, obtaining a time-series waveform of a frequency slope and a time-series waveform of frequency fluctuation, and applying frequency analysis to these waveforms. During the frequency analysis, a power spectrum of a frequency is obtained that corresponds to each of a functional adjustment signal, a fatigue reception signal, and an activity adjustment signal determined in advance is obtained. Then, the state of a human being is determined based on time-series change in each power spectrum. The fatigue reception signal indicates a degree of progress of fatigue in a usual active state. Thus, by comparing this degree with respective degrees of predominance of the functional adjustment signal and the activity adjustment signal as their distribution rates, the state of a human being (such as a relaxed state, a fatigued state, a state where sympathetic nerves are predominant, or a state where parasympathetic nerves are predominant) can be determined more precisely.
According to a technique disclosed in patent literature 2 by the applicant of this application, a biological signal extracted from an air cushion to support the lumbar and its vicinity of a human being is differentiated twice to catch a sleep prediction signal more precisely. According to a technique disclosed in Patent literature 3, a biological signal extracted from an air cushion to support the lumbar and its vicinity is also differentiated twice and an acceleration pulse wave aging index is obtained using a resultant second derivative waveform, thereby determining the presence or absence of alcohol drinking.